Promote, Suppress, Iterate: How Language Models Answer One-to-Many Factual Queries

• URL: http://arxiv.org/abs/2502.20475v2
• Date: Wed, 05 Mar 2025 13:22:47 GMT
• Title: Promote, Suppress, Iterate: How Language Models Answer One-to-Many Factual Queries
• Abstract summary: To answer one-to-many factual queries, a language model (LM) must simultaneously recall knowledge and avoid repeating previous answers.<n>Across multiple datasets and models, we identify a promote-then-suppress mechanism: the model first recalls all answers, and then suppresses previously generated ones.
• Score: 21.28307426740275
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: To answer one-to-many factual queries (e.g., listing cities of a country), a language model (LM) must simultaneously recall knowledge and avoid repeating previous answers. How are these two subtasks implemented and integrated internally? Across multiple datasets and models, we identify a promote-then-suppress mechanism: the model first recalls all answers, and then suppresses previously generated ones. Specifically, LMs use both the subject and previous answer tokens to perform knowledge recall, with attention propagating subject information and MLPs promoting the answers. Then, attention attends to and suppresses previous answer tokens, while MLPs amplify the suppression signal. Our mechanism is corroborated by extensive experimental evidence: in addition to using early decoding and causal tracing, we analyze how components use different tokens by introducing both Token Lens, which decodes aggregated attention updates from specified tokens, and a knockout method that analyzes changes in MLP outputs after removing attention to specified tokens. Overall, we provide new insights into how LMs' internal components interact with different input tokens to support complex factual recall. Code is available at https://github.com/Lorenayannnnn/how-lms-answer-one-to-many-factual-queries.

Related papers

• Token Assorted: Mixing Latent and Text Tokens for Improved Language Model Reasoning [53.57895922042783]
  Large Language Models (LLMs) excel at reasoning and planning when trained on chainof-thought (CoT) data.<n>We propose a hybrid representation of the reasoning process, where we partially abstract away the initial reasoning steps using latent discrete tokens.
  arXiv  Detail & Related papers  (2025-02-05T15:33:00Z)
• Harnessing Large Language Models for Knowledge Graph Question Answering via Adaptive Multi-Aspect Retrieval-Augmentation [81.18701211912779]
  We introduce an Adaptive Multi-Aspect Retrieval-augmented over KGs (Amar) framework.<n>This method retrieves knowledge including entities, relations, and subgraphs, and converts each piece of retrieved text into prompt embeddings.<n>Our method has achieved state-of-the-art performance on two common datasets.
  arXiv  Detail & Related papers  (2024-12-24T16:38:04Z)
• Answer, Assemble, Ace: Understanding How LMs Answer Multiple Choice Questions [103.20281438405111]
  Multiple-choice question answering (MCQA) is a key competence of performant transformer language models. We employ vocabulary projection and activation patching methods to localize key hidden states that encode relevant information for predicting the correct answer. We show that subsequent layers increase the probability of the predicted answer symbol in vocabulary space, and that this probability increase is associated with a sparse set of attention heads with unique roles.
  arXiv  Detail & Related papers  (2024-07-21T00:10:23Z)
• Integrating Large Language Models with Graph-based Reasoning for Conversational Question Answering [58.17090503446995]
  We focus on a conversational question answering task which combines the challenges of understanding questions in context and reasoning over evidence gathered from heterogeneous sources like text, knowledge graphs, tables, and infoboxes. Our method utilizes a graph structured representation to aggregate information about a question and its context.
  arXiv  Detail & Related papers  (2024-06-14T13:28:03Z)
• Where is the answer? Investigating Positional Bias in Language Model Knowledge Extraction [36.40833517478628]
  Large language models require updates to remain up-to-date or adapt to new domains. One key is memorizing the latest information in a way that the memorized information is extractable with a query prompt. Despite minimizing document perplexity during fine-tuning, LLMs struggle to extract information through a prompt sentence.
  arXiv  Detail & Related papers  (2024-02-16T06:29:16Z)
• Summing Up the Facts: Additive Mechanisms Behind Factual Recall in LLMs [1.5571776694273143]
  We focus on the most basic form of this task -- factual recall. We find that the mechanistic story behind factual recall is more complex than previously thought.
  arXiv  Detail & Related papers  (2024-02-11T22:58:49Z)
• Uncertainty Guided Global Memory Improves Multi-Hop Question Answering [3.7013865226473848]
  We propose a two-stage method that first collects relevant information over the entire document to the memory and then combines it with local context to solve the task. Our experimental results show that fine-tuning a pre-trained model with memory-augmented input, including the most certain global elements, improves the model's performance.
  arXiv  Detail & Related papers  (2023-11-29T23:45:57Z)
• OPERA: Alleviating Hallucination in Multi-Modal Large Language Models via Over-Trust Penalty and Retrospection-Allocation [124.9008419182485]
  We present OPERA, a novel MLLM decoding method grounded in an Over-trust Penalty and a Retrospection-Allocation strategy. Our approach begins with an interesting observation that, most hallucinations are closely tied to the knowledge aggregation patterns in the self-attention matrix. Based on the observation, OPERA introduces a penalty term on the model logits during the beam-search decoding to mitigate the over-trust issue.
  arXiv  Detail & Related papers  (2023-11-29T18:57:07Z)
• Explaining Interactions Between Text Spans [50.70253702800355]
  Reasoning over spans of tokens from different parts of the input is essential for natural language understanding. We introduce SpanEx, a dataset of human span interaction explanations for two NLU tasks: NLI and FC. We then investigate the decision-making processes of multiple fine-tuned large language models in terms of the employed connections between spans.
  arXiv  Detail & Related papers  (2023-10-20T13:52:37Z)
• Dissecting Recall of Factual Associations in Auto-Regressive Language Models [41.71388509750695]
  Transformer-based language models (LMs) are known to capture factual knowledge in their parameters. We study how the model aggregates information about the subject and relation to predict the correct attribute. Our findings introduce a comprehensive view of how factual associations are stored and extracted internally in LMs.
  arXiv  Detail & Related papers  (2023-04-28T11:26:17Z)
• Recitation-Augmented Language Models [85.30591349383849]
  We show that RECITE is a powerful paradigm for knowledge-intensive NLP tasks. Specifically, we show that by utilizing recitation as the intermediate step, a recite-and-answer scheme can achieve new state-of-the-art performance.
  arXiv  Detail & Related papers  (2022-10-04T00:49:20Z)
• Protum: A New Method For Prompt Tuning Based on "[MASK]" [12.057434751507552]
  We propose a new textbfPrompt textbfTuning based on "[textbfMASK]" (textbfProtum) method in this paper. Our textbfProtum can achieve much better performance than fine-tuning after continuous pre-training with less time consumption.
  arXiv  Detail & Related papers  (2022-01-28T13:34:30Z)
• SWAT: Spatial Structure Within and Among Tokens [53.525469741515884]
  We argue that models can have significant gains when spatial structure is preserved during tokenization. We propose two key contributions: (1) Structure-aware Tokenization and, (2) Structure-aware Mixing.
  arXiv  Detail & Related papers  (2021-11-26T18:59:38Z)
• Learning to Ask Conversational Questions by Optimizing Levenshtein Distance [83.53855889592734]
  We introduce a Reinforcement Iterative Sequence Editing (RISE) framework that optimize the minimum Levenshtein distance (MLD) through explicit editing actions. RISE is able to pay attention to tokens that are related to conversational characteristics. Experimental results on two benchmark datasets show that RISE significantly outperforms state-of-the-art methods.
  arXiv  Detail & Related papers  (2021-06-30T08:44:19Z)
• Answer Generation through Unified Memories over Multiple Passages [10.965065178451104]
  Our method is called neural answer Generation through Unified Memories over Multiple Passages (GUM-MP) First, it determines which tokens in the passages are matched to the question. In particular, it investigates matches between tokens in positive passages, which are assigned to the question, and those in negative passages, which are not related to the question. Finally, it encodes the token sequences with the above two matching results into unified memories in the passage encoders and learns the answer sequence by using an encoder-decoder with a multiple-pointer-generator mechanism.
  arXiv  Detail & Related papers  (2020-04-22T11:46:40Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.